Bayesian prediction of bacterial growth temperature range based on genome sequences

Area of Science:

• Genomics and Bioinformatics
• Enzyme Engineering
• Microbial Ecology

Background:

• Bacterial habitat preference correlates with enzyme production, particularly those stable at extreme temperatures.
• Predicting enzyme-producing capabilities from genomic sequences can streamline the search for industrially relevant microorganisms.
• Accurate genomic prediction reduces the need for extensive and time-consuming culturing experiments.

Purpose of the Study:

• To identify genomic features for predicting bacterial thermophilicity classes (thermophiles, mesophiles, psychrophiles).
• To compare the predictive power of protein families versus basic sequence features.
• To develop a computational tool for efficient thermophilicity prediction.

Main Methods:

• Identified 40 protein families discriminating between thermophilicity classes.
• Compared protein family performance against 87 basic sequence features using naïve Bayesian inference.
• Developed a dedicated computer program for thermophilicity prediction based on genomic data.

Main Results:

• A Matthews correlation coefficient of up to 0.68 was achieved in predicting optimal bacterial temperature ranges.
• Combining protein families and structural features yielded superior predictive performance compared to using either alone.
• The developed program effectively distinguishes between thermophilic, mesophilic, and psychrophilic bacterial genomes.

Conclusions:

• Protein families associated with specific thermophilicity classes are effective predictors of bacterial temperature adaptation.
• The naïve Bayesian approach demonstrates efficacy for thermophilicity prediction tasks.
• The developed computational tool efficiently classifies bacterial genomes based on their predicted optimal growth temperature.